Machine learning based triple region segmentation framework using level set on PACS

Abstract

Certain embodiments of the present invention provide methods and systems for triple region image segmentation. Certain embodiments provide a method for triple region image segmentation on a picture archiving and communication system. The method includes forming an initial contour for an image including three regions using principal component analysis and a support vector machine. The method also includes segmenting the image into three regions using a single level set function based on the initial contour. Certain embodiments provide an image processing system facilitating triple region segmentation of an image. The system includes a pattern classifier including a support vector machine, the pattern classifier forming an initial contour for an image including three regions using principal component analysis and the support vector machine. The system also includes a triple region segmenter segmenting the image into three regions using a single level set function based on the initial contour.

Description

RELATED APPLICATIONS[0001]This application claims priority to and benefit of U.S. Provisional Application No. 60 / 827,579, filed on Sep. 29, 2006, and U.S. Provisional Application No. 60 / 827,822, filed on Oct. 2, 2006, both of which are herein incorporated by reference in their entirety.FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT[0002][Not Applicable]MICROFICHE / COPYRIGHT REFERENCE[0003][Not Applicable]BACKGROUND OF THE INVENTION[0004]The present invention generally relates to medical image analysis. In particular, the present invention relates to use of image segmentation in computer-aided medical image analysis and diagnosis.[0005]Due to complexity of medical image segmentation and speed and accuracy requirements of a clinical environment, clinical image segmentation is a challenging topic in computer-aided medical image analysis and diagnosis.[0006]Segmentation is used, among other things, to distinguish objects from background. Techniques, such as threshold techniques, edge-based ...

AI Technical Summary

Benefits of technology

The method enables faster and more robust triple-region segmentation, reducing processing time and improving convergence, making it suitable for clinical applications and real-time use in medical imaging.

Problems solved by technology

Due to complexity of medical image segmentation and speed and accuracy requirements of a clinical environment, clinical image segmentation is a challenging topic in computer-aided medical image analysis and diagnosis.

Because spatial information is ignored, however, blurred region boundaries can create havoc for image segmentation.

However, edge-based methods exhibit weakness in connecting broken contour lines and thus may be prone to failure in the presence of blurring.

Overly stringent merging criteria may create fragmentation; lenient criterion may overlook blurred boundaries and result in over merging.

However, although some work has been reported on medical image segmentation using level set, many challenges remain in the application of level set for medical image segmentation, especially clinical image segmentation, even for triple-region images.

However, these approaches are time consuming and / or suffer from convergence problems.

Current level sets are not suitable for clinical segmentation.

Not only are current level set function techniques time consuming due to complicated medical structure, but level set functions are also sensitive to placement of an initial contour.

Therefore, the running time of current level set methods heavily relies on the position and size of the initial curves and the complexity of objects, for example.

As soon as more than two regions are considered, current level set solutions lose much of their attractiveness.

These techniques are of limited use and have significant limitations as described below.

Unfortunately, this approach results in empty regions if log2N is a floating point number rather than an integer.

While coupled level set functions have attempted to segment triple regions, coupled level set solutions suffer from slow convergence and local minimization problems.

Images